Yttrium-90 Radioembolization Dosimetry: What Trainees Need to Know

Yttrium-90 radioembolization (Y90-RE), also known as transarterial radioembolization (TARE) or selective internal radiation therapy (SIRT), is a form of brachytherapy that has become an established liver-directed therapy for primary and secondary hepatic malignancies. ^{1 2 3 4 5 6 7} While a degree of embolization and ischemia may occur, the dominant mechanism of action for Y90-RE is radiation-induced necrosis from targeted transarterial administration of millions of Y90-labeled microspheres. The Y90 within these microspheres exert their effects primarily by undergoing β-decay to stable zirconium-90, which is not known to have any clinical effects. ^{8 9 10} The β-decay of Y90 results in the release of high-energy β-particles (i.e., electrons or β−) with an average energy of 0.9267 MeV (maximum of 2.28 MeV) and a half-life of 64.04 hours (2.67 days), which translates to 94% of the Y90 radiation being delivered within 11 days. These β-particles penetrate nearby tissues an average of 2.5 mm (maximum of 11 mm), resulting in the sought-after effect of radiation damage to nearby structures. ¹¹ Additional types of radiation also occur as a result of Y90 decay. Although these are summarized in Fig. 1 , an in-depth discussion of them is beyond the scope of this article.

Fig. 1.

Yttrium-90 decay products and their clinical applications. Yttrium-90 predominantly undergoes β-decay to emit high-energy β-particles that are used clinically for targeted radiotherapy—which include direct injection of Y90 into a body cavity or space, conjugation of Y90 to an antibody for radioimmunotherapy (RIT), conjugation of Y90 to a peptide for peptide receptor radionuclide therapy (PRRT), or incorporation of Y90 to resin or glass microspheres for Y90 radioembolization (Y90-RE) therapy. As a result of the high-energy β-particle emission, a continuous spectrum of bremsstrahlung radiation occurs—which can be imaged using conventional nuclear medicine imaging systems (i.e., SPECT, SPECT/CT, planar gamma cameras). High-energy β-radiation also partakes in a phenomenon called Cherenkov radiation, which produces a continuous spectrum of ultraviolet and visible light photons (i.e., Cherenkov luminescence) which can be imaged using Cherenkov luminescence imaging (CLI). While β-decay is the predominant decay mechanism of Y90, every 32 per million Y90 decays result in an internal pair production (gamma-decay) that produces annihilation radiation that can be imaged using conventional PET/CT or PET/MRI systems.

Currently, there are two commercially available and Food and Drug Administration (FDA)-approved radioembolization microspheres in the United States: resin microspheres (SIR-Spheres; Sirtex Medical Inc, Woburn, MA), whose original formulations were developed in the mid-20th century, and glass microspheres (TheraSphere; Boston Scientific, Marlborough, MA), which were developed in the early 1980s. ⁹ Properties of these biocompatible and nonbiodegradable microspheres, at the time of calibration, are outlined in Table 1 . ^{9 12 13} Resin microspheres are FDA approved only for the treatment of unresectable metastatic liver tumors (MLTs) from primary colorectal cancer with adjuvant intrahepatic artery chemotherapy (IHAC) of FUDR (Floxuridine). ¹⁴ Glass microspheres are FDA approved, under a humanitarian drug exemption, only for the sole or neoadjuvant treatment of unresectable hepatocellular carcinoma (HCC). ¹⁵ Nevertheless, both types of microspheres are frequently used off-label for the treatment of various primary or secondary hepatic malignancies. ^{16 17}

Table 1. Properties of the commercially available glass and resin Y90 microspheres at the time of calibration.

	Glass	Resin
Isotope attachment	Incorporated into glass matrix	Attached to resin surface
Mean diameters (μm)	25	32.5 ± 2.5
Diameter range (μm)	20–30	20–60
Microspheres per vial	1.2 million for 3 GBq; 8 mil for 20 GBq	40–80 million
Available standard doses (GBq)	3, 5, 7, 10, 15, and 20	3
Specific activity (Bq per microsphere)	2,500	50
Specific gravity (g/dL)	3.6	1.6

Notes: Please check the latest package insert for updated information, including availability of customizable doses. For reference, the specific gravity of blood is 1.05 g/dL.

As a result of the mounting evidence for a clear dose–effect relationship, ¹⁰ the goal of Y90-RE has evolved to reflect a classical principle of oncology—which is to deliver the maximum tolerated dose. Achieving this goal requires understanding of the multiple steps in the pre-, peri-, and posttherapy phases of Y90-RE. Several articles have sought to comprehensively explain the rationale and technical challenges found in each of these steps. ^{18 19} However, a paucity of literature comprehensively describing the technical strengths and challenges of the commonly used Y90-RE dosimetry models remains. As an integral part of the team and an authorized user of Y90-RE devices, the interventional radiologist must have a fundamental understanding of the involved dosimetry. Therefore, the aim of this article is to provide a fundamental background of the rationale, limitations, and strengths involved in Y90-RE dosimetry planning, and the strategies employed in clinical practice when treating patients with Y90-RE.

Important Fundamentals in Dosimetry

Important Definitions

• Activity : Energy lost per unit time by an unstable atomic nucleus as a result of radioactive decay. Commonly measured as decays per second or Becquerel (Bq). A non-SI metric unit used in the United States is the Curie (Ci), which equals 37 GBq.

• Volume : The amount of tissue in which the isotope activity will be distributed in. Based on phantom and cadaver models, liver volume can be approximated to liver weight by using a conversion factor of 1.03 g/mL. ^{20 21} While densitovolumetry techniques can be used to personalize lung volumes to weight, a standard mass of 1,000 g is often assumed for lung dosimetry. ²²

• Absorbed dose : Amount of energy (J) deposited within a specific amount of mass (kg)—measured as gray (Gy). Since the mass of the lung and liver can be deduced from their volumes, the mean absorbed dose can often be practically simplified as a factor of the activity distributed over a specific volume.

• Compartments in dosimetry models : Dosimetry models are based on experiments originally performed on physical models. ²¹ Therefore, the number of “compartments” within a chosen dosimetry model literally translates to the number of areas (e.g., organs, tumors, normal tissue) that the dosimetry model mathematically considers. Important but often inaccurate assumptions for all clinically used dosimetry models are that (1) the delivered isotope is uniformly distributed within the targeted compartment(s) and (2) the isotope activity is fully absorbed within the tissue. For example, a single-compartment model considers only one area (i.e., if the targeted area contains both tumor and nontumor tissue, both are assumed to have equal homogenous distribution of Y90). In a multicompartment model, multiple areas are considered (i.e., if the targeted area contains both tumor and nontumor tissues, each tissue is assumed to have different amounts of homogenous Y90 distribution). In general, increasing the number of compartments within a model increases the accuracy, complexity, and tedium of the dosimetry model.

Goal of Therapy

The tailoring of Y90-RE therapy will depend on the overall clinical profile and the extent of tumor burden. In general, most patients can be categorized under one of two general intents of treatment, of which four accepted Y90-RE scenarios exist:

1. Curative intent:

   1. Y90 radiation segmentectomy (Y90-RS) : Application of ablative dose, as curative therapy or bridge/downstage to liver transplantation, for single lesions or lesions confined to ≤2 segments.

   2. Y90 radiation lobectomy (Y90-RL) : Application of an ablative dose, with intent of disease control, to induce treatment-related ipsilateral targeted lobe atrophy and subsequent contralateral lobe hypertrophy. Goal is to bridge to disease-ridden lobe resection or, in select cases, liver transplantation.

2. Palliative intent:

   1. Multifocal unilobar or bilobar disease is present : Application of personalized dose with intent of palliation or delaying disease progression. Concurrent and/or additional therapies are often considered.

   2. Presence of macrovascular tumor invasion, regardless of tumor size or distribution : Application of personalized dose with intent of palliation or delaying disease progression. Concurrent and/or additional therapies are often considered. Surgical conversion or downstaging is possible in select cases. ²³

Standardized Hepatic Anatomy Nomenclature

To facilitate collaboration with surgical colleagues, it is important to be familiar with the Brisbane nomenclature, which is considered the standard surgical liver anatomy and resection nomenclature schema. Introduced in 2000 by the terminology committee of the International Hepato-Pancreato-Biliary Association, the Brisbane nomenclature has three orders of division based on the liver's internal anatomy: hemilivers, sectors or sections, and segments. ²⁴ In the realm of radiology, however, the most commonly used nomenclature scheme for hepatic anatomy remains the Couinaud liver segmentation nomenclature, which includes lobes, livers, sectors, and segments ( Table 2 ). ²⁵ With the addition of “lobes” as the first order division, Couinaud's divisions are somewhat similar to Brisbane's nomenclature—with the alteration of the term “livers” being used instead of “hemilivers” and “sectors” being used instead of “sections.” In a manner often used by the interventional radiology community but discouraged by the Brisbane nomenclature, the liver is often separated into two lobes by its surface/visible falciform ligament. ¹⁸ However, this simple division can often result in confusion. For example, using Couinaud's divisions, the left lobe (segments II and III) does not equal the left liver (segments II, III, and IV). Nevertheless, when an interventional radiologist refers to having performed a left Y90-RL, they often refer to the treatment of segments II to IV (not just segments II and III). Therefore, when collaborating with other specialties and planning the appropriate dosimetry plans for Y90-RE therapy, it is important to understand the nuances of the commonly used nomenclatures.

Table 2. Terms and definitions of Couinaud's four orders of liver division.

Pretreatment Planning—Clinical, Imaging, and Angiographic Evaluation

In light of the several articles describing the pretreatment planning phases of Y90-RE, ^{8 18 26} this article will only focus on the dosimetry aspects of the pretreatment planning. Briefly, the goal of pretreatment planning is to reduce the extent of uncertainty that will be reflected in the radiation dosimetry model and expected outcome. In the clinical and imaging evaluation phase of planning, an important consideration is the relationship of the tumor characteristics and the patient's clinical profile—for this relationship will help guide the intent of Y90-RE therapy. Contouring of the targeted tissues is often done on preprocedural multiphasic CT or MRI by either a dedicated 3D Dosimetry Laboratory or nuclear medicine physician. Regarding the angiographic evaluation, an important consideration is to determine an optimal catheter position that permits the reproducible and appropriate delivery of the microspheres to the targeted areas—particularly in patients who have had prior therapies (e.g., transarterial chemoembolization, microwave ablation) that may affect the peri- and intratumoral vascular flow.

⁹⁹ mTc Macroaggregated Albumin Imaging

Because of the cancer-associated angiogenesis, a certain degree of vascular caliber heterogeneity is to be expected within the targeted tumor tissues. Therefore, a percentage of the delivered Y90 microspheres will be unlikely to lodge within the targeted arterial tumor vasculature and thus progress (i.e., shunt) into the next major capillary system in the lungs. Should this process occur in significant quantity, this hepatopulmonary shunting will result in the inadvertent diffuse delivery of Y90 radiation to the bilateral lung parenchyma—which may cause a pathologic response known as radiation-induced pneumonitis (RP). ^{27 28 29 30}

Current standards of care dictate that a surrogate radiopharmaceutical, Technetium ⁹⁹ mTc macro-aggregated albumin (Tc-MAA), be used to predict the degree of arteriovenous shunting as a metric called the lung-shunt fraction (LSF). What makes Tc-MAA the currently preferred standard surrogate for the distribution of Y90 particles is the low cost, widespread availability, comparable size to glass and resin Y90 particles, and the ability for albumin aggregates to be sufficiently fragile for the mechanical capillary micro-occlusion to only be temporary (i.e., susceptible to the continued erosion and fragmentation by the blood flow). Of note, the Tc-MAA shunt study has repeatedly been found to be a poor surrogate of Y90 distribution and absorbed dose within HCC tumors for a variety of reasons. ^{31 32}

Given the short physical half-life of 6 hours and the concurrent biodegradation of the Tc-MAA, it is recommended that Tc-MAA imaging occur within 60 minutes post injection. ³³ Image capture of the Tc-MAA deposition is commonly attained with planar imaging. However, multiple studies have suggested that tomographic imaging, such as single-photon emission computed tomography/CT (SPECT/CT), can provide a more accurate representation of the Tc-MAA biodistribution—especially when extrahepatic deposition or nonhomogeneous Tc-MAA distribution occurs. ³⁴ Significant limitations in both planar and tomographic imaging modalities exist and are discussed elsewhere. ³⁵

The LSF is quantitatively estimated as the ratio of activity in the lungs divided by the summed activity in the liver and lungs, and is exemplified mathematically as follows:

where LSF is the lung shunt fraction as a percentage, A _lung is the activity within the lung, and A _liver is the activity within the total hepatic parenchyma.

Current interventional radiology guidelines used for limiting the incidence of RP after Y90-RL vary per type of microsphere—with resin microspheres being contraindicated in patients with a LSF greater than 20%, and glass microspheres being contraindicated in cases where the to-be-delivered single-treatment lung dose is expected to be greater than 30 Gy. ^{13 36} All Y90-RE cases, whether using resin or glass microspheres, should ensure that no more than 30 Gy per single-session or 50 Gy for life-time cumulative dose is ever absorbed by the bilateral lungs. ^{19 26} Of note, these maximum lung dose recommendations are based on limited studies from the late 1990s, ^{37 38} whose dosimetry limits were originally extrapolated from mid-20th century radiation oncology data of whole-lung external-beam radiation therapy for metastatic lung cancer. ³⁹ Although there is a paucity of more recent data exploring the incidence of RP after Y90-RE, some studies suggest that RP occurs less commonly than previously believed. ⁴⁰ Overall, knowing the LSF and the dose already absorbed by the lungs is essential to dosimetry planning and calculations, for it can be used as a correction factor for the required activity to be delivered to the target tissue.

Dosimetry Models for Calculation of Y90 Activity Prescription

To date, four dosimetry models have been developed and clinically used for the calculation of Y90 prescription activity. These are the empiric, body surface area (BSA), medical internal radiation dose (MIRD), and the partition dosimetry models. The empiric model has since been abandoned given safety concerns related to radiation-induced side effects. ² The clinical implications and history of the empiric model have been reviewed by Bilbao and Reiser. ⁴¹ The manufacturer of glass microspheres recommends the usage of the MIRD or the partition dosimetry model, ³⁶ while the manufacturer of resin microspheres recommends the usage of the BSA or the partition dosimetry model. ¹³

Medical Internal Radiation Dose Dosimetry Model

As part of an endeavor to help the nuclear medicine community with the preparation of absorbed dose estimates for radiopharmaceuticals, the MIRD Committee of the Society of Nuclear Medicine started developing the MIRD Pamphlets in 1968. ⁴² The current iteration of the MIRD method for liver-directed Y90-RE dosimetry was formulated in 1975. ⁴³ To truly understand the limitations of the MIRD model and the dosimetry models derived from it, one must first understand how the actual MIRD formula was created. Considering the existing articles covering the derivation of the MIRD equation in detail, ^{44 45} the derivation of the MIRD formula will be briefly summarized, starting with the generic equation for absorbed dose rate:

where k is a constant that converts the dose rate to a desired unit, E is the average energy emitted per nuclear transition, A is the source activity, and m is the mass of the tissue that the radiation is absorbed within. Since the MIRD formula assumes that the distribution is uniform and permanent within a single compartment (i.e., real-world biological clearance or heterogenous distribution of Y90 is ignored), only the decay in activity that occurs as a result of isotope decay is considered. The generic activity of an isotope at a certain time, A(t) , can be represented as:

where A ₀ is the isotope activity at the time of calibration, T is the time since calibration, and t is the half-life of the radioactive isotope. Since the isotope is assumed to completely decay within the tissues, Eq. 3 can be integrated to infinity and implemented to Eq. 2 to attain the absorbed dose (D).

To make this generic absorbed dose equation applicable to Y90-RE, the Y90 isotope characteristics (energy released: 0.9267 MeV [Bq/s] and half-life: 64.04 hours) must be put into Eq. 4. The resulting equation, with the k constant adjusted to provide an absorbed dose in Gy as a function of GBq per kg, is

For simplicity, the number 49.38 is often rounded up to 50. Of note, Eq. 5 can be used to calculate the Y90 dose delivered to any targeted compartment, including the lung or tumor tissue.

Since the interventional radiologist usually knows the dose desired to be given and the mass (i.e., volume) of the targeted area, Eq. 5 can be rewritten to provide the needed Y90 activity to be administered at the time of Y90-RE therapy.

To account for the percentage of activity that will shunt to the lungs, the LSF correction is implemented to the formula. Assuming that 100% of the Y90 dose is injected into the patient, the following formula represents the official MIRD equation that is used clinically:

Since the MIRD method considers only one compartment in its equation, it is a single-compartment model, of which its largest limitation is its assumption of uniform activity distribution within the targeted tissue (which often contains both tumor and nontumor tissue). In real practice, however, microspheres are always heterogeneously distributed. Hence, the MIRD method should be used cautiously in settings where the activity biodistribution, as visualized in Tc-MAA imaging, is highly heterogenous. Such a situation may occur in very large tumors or in tumors that have been partially treated. Additionally, the derivation of the actual mathematical MIRD equation (Eq. 6) undertakes multiple “ideal situation” assumptions. Therefore, if the liver or lung tissue deviates significantly from the ideal density or shape assumed in the mathematical modeling (such as in the setting of severe cirrhosis or emphysema), careful usage of the method should be employed. While the MIRD method is frequently used clinically in prescribed activity calculation for glass-based Y90 microspheres, it too can be used off-label for resin-based Y90 microspheres. ⁴⁴

Partition Model

Endeavoring to address the MIRD model's limitation of uniform distribution within the targeted tissue, Ho et al implemented the tumor, nontumor, and lung tissue compartments into the MIRD equation. ⁴⁶ This three-compartment model, known as the partition model, was created in 1996, and is considered the most accurate and complex method for the calculation of prescribed Y90 activity. ^{18 19} Understanding the limitations of the model will rest upon the grasp of its major assumptions.

In a similar fashion as the MIRD model, the interventional radiologist will inject a certain amount of activity upon a specified target area. What makes the partition model more accurate, however, is that it separates the total activity ( A _Total ) being given to the targeted area into two compartments: tumor ( A _T ) and nontumor ( A _N ) tissue. Mathematically, this is exemplified as:

Fundamentally, partition model dosimetry is all about solving for A _T and A _N within Eq. 8. To do so, however, multiple assumptions must be made. The most important assumption is that the partition model assumes that each tumor and nontumor tissue compartment will have their own homogenous Y90 distributions within themselves. Therefore, it assumes that the relationship of the absorbed dose between the two compartments is cleanly linear. This relationship is exemplified within the partition model equation by a dosimetric parameter called the R _T/N ratio—which is a unitless ratio that is mathematically defined as the ratio between the absorbed doses in the tumor ( D _T ) and nontumor ( D _N ) compartments. ⁴⁶

The fundamental concept of the R _T/N ratio is that the linear relationship elucidated by it will help us calculate the highest dose able to be given to the tumor compartment ( D _T ) as a factor of the maximum dose acceptable to the nontumor compartment ( D _N ). In other words, the desired D _N (and by extension the desired A _N ) will dictate the D _T (and by extension the A _T ). With this assumption in mind, the R _T/N ratio can be used to help solve for A _Total in Eq. 8. However, to do this, the R _T/N ratio must first be derived down to a form that is clinically and mathematically usable.

From Eq. 5, it is known that the absorbed dose is proportional to the activity within a specified mass. Therefore, Eq. 5 can be used to deduce that the absorbed dose ratio between the tumor and nontumor compartments is:

where A _T is the deposited activity within the tumor compartment, m _N is the mass of the nontumor compartment, A _N is the deposited activity within the nontumor compartment, and m _T is the mass of the tumor compartment.

Since decay counts are innately proportional to the deposited isotope's activity, it is assumed that the decay counts (i.e., “uptake counts”) captured within the contoured tumor ( C _T ) and nontumor ( C _N ) compartments during Tc-MAA or Y90 Bremsstrahlung imaging will too be proportional, but not equal, to the deposited activity within the contoured tumor ( A _T ) and nontumor ( A _N ) compartments. Therefore, the assumed linear relationship (i.e., ratio) in counts and activity between the tumor and nontumor compartments can be used to mathematically exploit the relationship between counts and activity. This relationship between the ratio of counts and the ratio of activity is exemplified as follows:

Incorporating Eq. 11 into Eq. 10 results in the clinically used R _T/N ratio equation:

Please note that Eq. 11 and Eq. 12 do not suggest that counts are equal to activity, but instead that the ratio of counts is equal to the ratio of the activity.

Using the clinically usable form of the R _T/N ratio in Eq. 12, one can solve for A _Total in Eq. 8 by first rearranging the R _T/N ratio variables. Specifically, Eq. 12 must be rearranged so that the activity in the tumor ( A _T ) compartment will be proportional to the activity in the nontumor ( A _N ) compartment.

Since it is known from Eq. 5 that the absorbed dose is proportional to the activity within a specified mass, Eq. 5 can be implemented into Eq. 13 to deduct the tumor activity ( A _T ) as a factor of the absorbed dose in the nontumor ( D _N ) compartment.

Using Eq. 14 and Eq. 6 set up to represent A _N as factor of D _N , one can start solving for A _Total in Eq. 8.

Rearranging Eq. 15, the following equation is attained:

wherein the total activity to be delivered within the targeted area ( A _Total ) is in GBq, the maximum acceptable dose desired within the nontumor tissue ( D _N ) is in Gy, and the tumor ( m _T ) and nontumor ( m _N ) masses are in kg.

To account for the percentage of activity that will shunt to the lungs, the LSF correction is implemented to the formula. Assuming that 100% of the Y90 dose is injected into the patient, the following formula represents the official partition model equation that is used clinically:

Since the partition model equation considers three compartments, it is a three-compartment model, of which its largest limitation is the assumption that the R _T/N ratio attained from the Tc-MAA imaging will exactly be the same as in the Y90-RE therapy. While the catheter position in which the dose will be injected should remain the same in both the Tc-MAA study and Y90-RE therapy, it is not always feasible to do so in real practice. Furthermore, significant amount of activity acquisition variability exists within the Tc-MAA imaging portions of the R _T/N ratio calculation, with some research groups advocating for the implementation of improved R _T/N ratio calculation methods. ^{47 48 49} Since the partition model is derived from the MIRD method, it too shares all the assumptions detailed in the prior MIRD model section. The partition model can be used for both resin- and glass-based Y90 microspheres. ⁵⁰

Body Surface Area Model

With the advent of a total liver volume formula derived from BSA, particularly in western physique patients, ⁵¹ a modification to the empiric method was proposed in the early 2000s so that the activity to be delivered to a lobe could be easily calculated. ⁴¹ This new method was called the “BSA model,” and because of its simplicity and semiempirical nature, it remains a popular method for the dosimetry calculation for resin microspheres.

The BSA method semiempirically calculates the prescribed activity for each patient using only the BSA formula and the tumor burden within a targeted tissue. ^{8 52}

where A [GBq] is the to be prescribed activity within the targeted tissue, V _T is the tumor volume, V _N is the normal tissue volume, and BSA is the result of the Du Bois BSA formula shown below:

Since the BSA model considers only one compartment in its equation, it is a single-compartment model, of which its largest limitation is its semiempirical nature. Like the now-retired empirical model, the BSA model can separately and semiempirically consider the LSF to reduce the risk of receiving unwanted radiation within the lungs (e.g., 20 or 40% dose reduction for 10–15% or 15–20% LSF, respectively). Nevertheless, the limited personalization of the BSA model leaves it at a high risk for either undertreating or overtreating patients. ¹⁸ Furthermore, its single-compartment model, its disregard for the R _T/N ratio, and its inability to implement a desired absorbed dose prior to the Y90-RE procedure leave the BSA method's dosimetry capabilities much to be desired. Currently, the BSA method is the only recommended method for Y90-RE treatment with resin-based microspheres. ¹³

How to Do Y90-RE Dosimetry

Determination of Target Dose

To make use of the dosimetry models, the interventional radiologist must decide on the dose to be delivered to the targeted area(s). Part of this decision is based on the goal of therapy for each individual patient, for which it is based on the tumor characteristics, hepatic reserve, and the patient's overall clinical profile. Once the goal of therapy is decided, the method of Y90-RE will need to be selected. The target dose, per type of Y90-RE therapy, is discussed below:

Y90 Radiation Segmentectomy

Curative-intent therapy that delivers an ablative dose to a targeted area containing tumor(s) expanding a maximum of ≤ 2 adjacent hepatic segments . ^{19 53} Since the concept was first described in 2011, ⁵³ most studies evaluating the effectiveness of Y90-RS have been performed using glass microspheres—and they recommend a targeted tumor dose of at least greater than 190 Gy. ¹⁹ Lately, resin microspheres too have been reported to be used for Y90-RS. ⁵⁴ However, to date this practice remains much less developed than that of glass Y90-RS. Nevertheless, select studies have recommended that a targeted tumor dose not to exceed 200 Gy when using resin microspheres. ⁵⁵ In general, since the tumor and targeted area are relatively small, the assumption is made that the whole targeted area will be “ablated” equally by a homogenous distribution of Y90 microspheres. Therefore, the MIRD model is often used for Y90-RS. The partition model can technically be used as well, especially in situations when the Tc-MAA imaging demonstrates a heterogenous Y90 biodistribution or, in our experience, when the R _T/N ratio is greater than 2 (which would suggest that an ablative tumor dose [D _T ] can be achieved with a reduced normal tissue dose [D _N ]). Especial consideration is needed when the spatial resolution limitations of the Tc-MAA imaging make the use of the partition model difficult (such as in the setting of tumors measuring ≤2 cm or with ill-defined margins). ³² Since the risk of toxicity is mitigated by the relatively small liver volume that is ablated, nearly all of the adverse events (fatigue, abdominal pain, nausea, fever) that may occur are expected to be mild and transient. ⁵⁶ Treatment of two tumors in two separate segments may be pursued in patients with good hepatic function reserve. ⁵⁷ Lastly, tumors abutting the colon, gallbladder, or stomach are safe to treat with Y90-RS. ⁵⁸

Y90 Radiation Lobectomy

Curative-intent treatment for patients with large unilobar tumors and who fit in the following general categories : (1) potential candidates to undergo curative surgical resection if sufficient future liver remnant is present after Y90-RL, ⁵⁹ (2) potential candidates for liver transplantation if within transplant criteria after Y90-RL, ¹⁹ or (3) not a potential surgical candidate secondary to significant portal hypertension, but with goal of therapy to cure the tumor and for the unaffected lobe to hypertrophy and maintain adequate hepatic function. ⁶⁰ The overall goal of Y90-RL is to deliver an ablative dose able to provide a tumoricidal effect, atrophy of the treated hepatic lobe (secondary to radiation-induced fibrosis and tissue retraction), and consequential hypertrophy of the contralateral lobe (secondary to gradual increase in redirected portal flow). ⁶⁰ The concept of Y90-RL was first described in 2009 using both glass and resin microspheres with the MIRD and BSA models, respectively. ^{61 62} Since then, the effectiveness of Y90-RL with resin and glass microspheres has been further evaluated—with all three dosimetry models demonstrating effectiveness. ⁵⁵ In Child–Pugh A patients undergoing Y90-RL with glass microspheres, contralateral liver hypertrophy has been found to occur when the targeted liver tissue receives a dose of 140 to 150 Gy using the MIRD model, or greater than 88 Gy for the normal liver tissue (D _N ) using the partition model. ⁶³ For resin microspheres, exact dosage recommendations for Y90-RL are limited—with many of the major studies demonstrating successful usage of the semiempirical BSA model (with few to no clinically significant adverse events reported). ^{64 65 66 67 68} The manufacturer's training manual recommends a maximal dose of 80 or 70 Gy, respectively, for noncirrhotic or cirrhotic normal liver tissue (D _N ) using the partition model. ⁶⁹ Despite reports that the normal liver is able to tolerate higher doses of Y90 resin microspheres, ⁷⁰ data on post-Y90-RE liver toxicity and its association with contralateral lobe hypertrophy remain somewhat limited ^{71 72} —suggesting that the maximum doses recommended by the manufacturer may be as adequate as other recommendation for Y90-RL with resin microspheres. If Y90-RS is technically feasible but Y90-RL is still the plan (which would imply that contralateral lobar hypertrophy is still desired), one can consider a “modified” Y90-RL, which is where a single-session ablative Y90-RS is followed by Y90-RL. ⁷³ This modified Y90-RL should be favored when technically and clinically feasible, for it provides the benefit of tumor control regardless of the patient's surgical outcome. Generally, at least 3 to 6 months should be allowed for the contralateral lobe to hypertrophy, with expected volumetric increases reaching 26 to 47%. ⁷⁴ Longer wait times are permissible as long as there is acceptable tumor control. ¹⁹ Portal vein embolization after lack of hypertrophy (and vice versa) remains currently investigational. Lastly, Y90-RL has been shown to be safe in patients with PVT, with some studies demonstrating greater amount of contralateral hypertrophy with glass Y90-RL in this patient population. ⁶⁰

Lobar Y90-RE (Non-Lobectomy)

Palliative-intent treatment, with the aim of diffuse tumor therapy to slow down the overall tumor progression . This strategy is usually reserved for patients with multilobar or multifocal tumors beyond the capabilities of curative therapies or potential liver transplantation. Dosimetry model to be used must be tailored to the tumor characteristics, hepatic reserve, and clinical picture. In the setting of poor hepatic reserve or a relatively “straightforward” tumor distribution, the partition model may afford a higher tumor dose while minimizing the dose to nontumoral tissue. If the tumor distribution is complex, a single-compartment dosimetry model, such as the BSA or MIRD model, may make the planning easier. If the MIRD model is used for glass microspheres, a mean absorbed dose of 120 Gy (range: 80–150 Gy) to the target area is recommended. ^{18 19 26 75} If the BSA method is used, careful attention must be paid to the patient's weight and height to ensure that the activity delivered is not at an unreasonable higher risk of causing liver toxicity. Since many of the patients undergoing Y90-RE lobar treatment have limited functional liver reserve, it is imperative to deliver a safe and effective dose in a manner that hypothetically permits the nontargeted liver tissue to solely carry the liver function should the targeted liver tissue be significantly injured. Therefore, consideration for personalized partition dosimetry, conservative dose delivery, and/or sequential unilateral lobe treatment should be considered whenever feasible.

Y90-RE in the Setting of Macrovascular Invasion

Depending on the location and extent of the macrovascular invasion, the intent can be either palliative or curative. Often, however, the presence of macrovascular invasion portends a poor prognosis and hence, treatment is more commonly palliative in nature. Nevertheless, Y90-RE has shown promising efficacy in this setting, ⁷⁶ with personalized dosimetry methods able to deliver high tumor doses demonstrating the most promising results. ⁷⁷ During the Tc-MAA shunt study, careful attention should be given to the catheter position, which should be proximal enough to perfuse all feeding vessels into both the tumor and tumor thrombus. During the dosimetry planning portion, a normal tissue dose ( D _N ) able to provide at least a greater than 205 Gy tumor dose ( D _T ) (i.e., treatment intensification) can be pursued with glass microspheres and the partition model. ⁷⁷ If Tc-MAA imaging demonstrates poor tumor perfusion or if tumor dose ( D _T ) is later found to be less than 205 Gy in post–glass Y90-RE dosimetry evaluation, concurrent and/or additional therapies should be highly considered. Should the MIRD model be used, a similar dosimetry approach as Y90-RL or lobar Y90-RE should be pursued as appropriate to the tumor and clinical picture. For resin microspheres, many studies have demonstrated the successful usage of the BSA model for Y90-RE in HCC patients with PVT. ^{78 79} To our knowledge, there are no published dose recommendations for the usage of resin microspheres with the partition model in the setting of HCC with macrovascular invasion. Whichever the case, judicious caution should be observed in cases where there is bilobar disease and macrovascular invasion, for these patients may have a fragile liver function and limited hepatic vascular flow reserve.

Y90-RE Retreatment

Defined as more than one radiation treatment to a previously targeted tissue, Y90-RE retreatment may be pursued in cases where the original therapy was not able to achieve a desired effect on the targeted tissue. A major concern with Y90-RE retreatment is the development of radioembolization-induced liver disease (REILD), a form of acute hepatic failure that usually arises within 2 months of retreatment as a result of radiation-induced necrosis of the functional hepatic parenchyma. ^{80 81 82} Risk factors for REILD are not fully understood, but they have been postulated to be associated with prior intra-arterial treatments (especially in more than two Y90-RE therapies), neoadjuvant or adjuvant chemotherapy, baseline cirrhosis, baseline model for end-stage liver disease (MELD) score greater than 8, and an elevated total administered dose. ^{80 81 83 84 85 86 87}

Regarding dosimetry planning in the setting of Y90-RE retreatment, judicious attention should be given to the cumulative absorbed dose to the lungs and nontumor liver tissue. Although no specific lifetime cumulative dose limit has been established for nontumor liver tissue, ^{8 82} personalized dose planning is recommended to deliver the highest tolerable tumor dose while minimizing the dose to nontumoral tissue. Additional advanced vascular techniques (e.g., coiling, ⁸⁸ absorbable gelatin powder, ⁸⁹ vascular plugs, ⁹⁰ balloon occlusion microcatheters) ⁹¹ may be used to modify the vascular flow to theoretically augment the dose delivered to the targeted tissue while minimizing nontargeted Y90 deposition. However, the impact of these techniques on dosimetry planning and survival outcomes remains under study.

Tailoring the Specific Activity of Glass and Resin Microspheres

The endeavor to maximize the dose–effect relationship at times is hampered by the challenges associated with attaining proper tumor coverage. Fortunately, achieving this feat has become easier with the advent of flexible dose shipping options for glass and resin microspheres. ^{92 93} As a result of the manufacturing process, ⁹ the specific activity of glass microspheres, at the time of calibration, is higher than that of resin microspheres ( Table 1 ). For glass microspheres, the manufacturer reports a shelf life of 12 days from calibration time—which is set to always occur at 12 pm eastern time on a Sunday. The commercially available shipping options in the United States are divided into two: week 1 (from calibration day, i.e., day 0, to day 7) and week 2 (from day 8 to day 12). ⁹³ In other words, the manufacturer of glass microspheres offers the option to achieve a desired Y90 activity with a greater number of lower specific activity Y90 glass microspheres. For resin microspheres, the manufacturer reports a shelf life of 24 hours from calibration—which is set to occur at 6 pm eastern time on the scheduled day of Y90-RE therapy. ⁹² The commercially available shipping options in the United States are divided in a per day basis, with the earliest option being up to 3 days precalibration (starting at 8 am of day 3) to 1 day postcalibration (ending at 6 pm of day 1). In other words, the manufacturer of resin microspheres offers more options to achieve a desired Y90 activity with a lesser number of higher specific activity Y90 resin microspheres than with a greater number of lower specific activity Y90 resin microspheres. Effectively, these flexible dose shipping options permit the customer a sizeable number of microsphere quantity and activity permutations able to address nearly all types of foreseeable clinical scenarios. Choosing the right permutation, however, can be more of an art than a science, given the relatively limited amount of literature on the topic (especially for resin microspheres). Nevertheless, the general idea is that large masses will theoretically benefit from a higher number of lower specific activity microspheres able to provide a more even biodistribution, while smaller masses will theoretically benefit from a lower number of higher specific activity microspheres. To further clarify this concept, the many permutations able to theoretically attain an activity of 3 GBq at the time of Y90 delivery are pictured in Fig. 2 .

Fig. 2.

Example of the many Y90 microsphere number and specific activity permutations able to provide 3 GBq of activity at the time of Y90 delivery. For the general clarification of the flexible dose concept, the relative relationship of the permutations to the targeted tumor size is pictured as well. Data shown were extrapolated from manufacturer's Y90 decay information. ^{92 93}

Dosimetry Pearls

1. Current iterations of resin and glass microspheres have been found to have different hepatic toxicity thresholds—postulated to be different as a result of their distinct physical properties and the manner in which they are supplied. ^{94 95} As such, there are certain circumstances in which one type of microsphere may theoretically be better than the other. However, while each Y90 microsphere may have their own theoretical advantages and disadvantages, to date there have been no comprehensive clinical trials demonstrating the superiority or inferiority of one over the other. ⁹

2. All dosimetry models were developed with numerous mathematical and physical model assumptions. Therefore, judicious attention must be given to cases that are “not straightforward” (i.e., prior treatments, complex tumor burden, and extreme height/weight).

3. Tc-MAA and Y90-RE imaging modalities have temporal and spatial resolution limitations that should always be considered when doing dosimetry planning or dose delivery assessment. Additionally, there is increasing evidence that Tc-MAA may be a poor predictor of Y90 biodistribution. ^{31 32}

4. Obtaining three-dimensional cone beam CT (CBCT) during angiographic mapping can provide vital information on the relative amount of tumor coverage provided by the to-be-intended microcatheter position. As such, tumor coverage using contrast on CBCT should be compared with the Tc-MAA distribution on SPECT/CT images to confirm that the microcatheter position where the MAA was administered will provide an effective and safe position for the administration of the Y90 microspheres.

Conclusion

Y90-RE is a multifaceted and potentially complex therapy that requires multidisciplinary collaboration for the safe and efficacious delivery of Y90 microspheres. It has the potential to replace or augment therapeutic options in patients with primary or secondary liver cancer. Understanding the origin, development, and usage of the Y90-RE dosimetry methods is imperative for the full understanding of the Y90-RE capabilities and limitations. Given that the goal of Y90-RE is to deliver the maximum tolerated dose able to maximize the dose–effect relationship, increased consideration should be given to multicompartmental dosimetry methods better able to provide personalized dosages. As further innovation, standardization, and collaborative efforts take place in the field of Y90-RE, the use of personalized, safe, and efficacious therapies is likely to be expected from the interventional radiology community.

Funding Statement

Funding/Support No funding was received to prepare this manuscript.

Erratum: This article has been corrected in accordance with the Erratum published on January 13, 2021. The arrows within equations 10, 12 and 14 have been corrected.